3 ms·

Interesting.. theoretically SSL should prevent MITM of this kind. But if you are able to redirect at DNS level, probably you can deliver your own certificate an

Interesting.. theoretically SSL should prevent MITM of this kind. But if you are able to redirect at DNS level, probably you can deliver your own certificate and its authority.

niels_olson 14y ago
I don't see how DNS MiTM is going to subvert the basic math of public key crypto. Even if the browser is implementing its own SSL stack, it would have access to validate those keys against those that ship with the OS. This SSL pinning seems to be something outside the crypto mechanism. It's flair.
- option_greek 14y ago
  You are correct. That should handle it.
gojomo 14y ago
Creating a phony certificate is much harder -- or at least by design it's supposed to be much harder -- than DNS redirection/spoofing. So there's not a "if you can do one you can probably do the other" kind of threat-equality. However, given how many Certificate Authorities have been flaky, compromised, or deferential to certain unsavory authorities, there probably are some actors who could launch both kinds of attacks. I expect both Google and Microsoft use some private mechanism for authenticating their updates, perhaps SSL but not dependent on any Certificate Authorities or secrets other than those they manage themselves. Pressure from some state authorities, or catastrophic failures in these companies' own security, could give some other malicious actor the power to hijack these updates. But that should be again harder than either DNS subversion or finding/compromising the weakest browser-trusted CA. (When Google suddenly became so mad at China a while back, my favorite theory -- wholly unsupported by any confirmed info, however -- was that Google had observed China's incursions trying to win control of the various Google auto-update mechanisms. Third-party malware injection via those avenues might threaten Google's very existence, which would justify the strongest possible reaction, in the name of self-preservation.)

Interesting.. theoretically SSL should prevent MITM of this kind. But if you are able to redirect at DNS level, probably you can deliver your own certificate and its authority.

niels_olson 14y ago

I don't see how DNS MiTM is going to subvert the basic math of public key crypto. Even if the browser is implementing its own SSL stack, it would have access to validate those keys against those that ship with the OS. This SSL pinning seems to be something outside the crypto mechanism. It's flair.

option_greek 14y ago
You are correct. That should handle it.

gojomo 14y ago

Creating a phony certificate is much harder -- or at least by design it's supposed to be much harder -- than DNS redirection/spoofing. So there's not a "if you can do one you can probably do the other" kind of threat-equality. However, given how many Certificate Authorities have been flaky, compromised, or deferential to certain unsavory authorities, there probably are some actors who could launch both kinds of attacks. I expect both Google and Microsoft use some private mechanism for authenticating their updates, perhaps SSL but not dependent on any Certificate Authorities or secrets other than those they manage themselves. Pressure from some state authorities, or catastrophic failures in these companies' own security, could give some other malicious actor the power to hijack these updates. But that should be again harder than either DNS subversion or finding/compromising the weakest browser-trusted CA. (When Google suddenly became so mad at China a while back, my favorite theory -- wholly unsupported by any confirmed info, however -- was that Google had observed China's incursions trying to win control of the various Google auto-update mechanisms. Third-party malware injection via those avenues might threaten Google's very existence, which would justify the strongest possible reaction, in the name of self-preservation.)